dsh-origin-bridge 0.4.1 → 0.4.3

package/asciiio.py

@@ -25,15 +25,59 @@ def _is_number(token):
     return bool(t) and bool(_NUM.match(t))
 
 
+_NAME_KEYWORDS = ("long name", "longname", "name", "display", "l")
+_UNIT_KEYWORDS = ("units", "unit", "u")
+# "Potential/V", "电流 (mA)", "Z'/ohm": instruments pack the unit into the name.
+_NAME_UNIT = re.compile(r"^(?P<name>[^/\(\[]+?)\s*(?:/\s*|\(|\[)\s*(?P<units>[^)\]]+?)\s*[\)\]]?$")
+_PAIR = re.compile(r"^\s*(?P<key>[^=:]{2,40}?)\s*[=:]\s*(?P<value>.+?)\s*$")
+
+
+def _row_cells(row, width):
+    return [str(row[c] if c < len(row) else "").strip() for c in range(width)]
+
+
+def _all_names(row, width):
+    cells = _row_cells(row, width)
+    return bool(cells) and all(cells) and not any(_is_number(c) for c in cells)
+
+
+def _units_like(row, width):
+    """A units row is short text ("s", "mV", "-") with no name-like slashes."""
+    cells = [c for c in _row_cells(row, width) if c]
+    return bool(cells) and not any(_is_number(c) for c in cells) \
+        and all(len(c) <= 6 and "/" not in c and "(" not in c for c in cells)
+
+
+def split_name_units(text):
+    m = _NAME_UNIT.match(str(text).strip())
+    if not m:
+        return str(text).strip(), ""
+    return m.group("name").strip(), m.group("units").strip()
+
+
 def _pick_name_row(head, width):
-    """The long-name row is the one with the most distinct tokens.
+    """Which header row carries the column long names?
 
-    Instrument dumps often start with a repeated filler row ("Sample Sample
-    Sample") or a caption, so taking head[0] blindly names every column 'Sample'.
+    Origin's own exports use a labelled block (Long Name / Units / Comments), so
+    that row wins. Instrument dumps (CH Instruments .txt and friends) put a page
+    of settings first and the column line immediately above the data, so the last
+    all-text row is the name row. Taking head[0] blindly named every column
+    "Jan. 31" / 2011 15:12:13 on those files.
     """
+    for i, row in enumerate(head):
+        cells = _row_cells(row, width)
+        if cells and cells[0].lower() in _NAME_KEYWORDS and _all_names(row, width):
+            return i
+    if head and _all_names(head[-1], width):
+        # "Time Signal Flag" followed by "s mV -": the short row is the units row.
+        if len(head) >= 2 and _all_names(head[-2], width) and _units_like(head[-1], width):
+            return len(head) - 2
+        return len(head) - 1
+    if len(head) >= 2 and _all_names(head[-2], width) and _units_like(head[-1], width):
+        return len(head) - 2
     best_i, best_score = 0, None
     for i, row in enumerate(head):
-        toks = [row[c].strip() for c in range(width)]
+        toks = _row_cells(row, width)
         uniq = len({t for t in toks if t})
         score = uniq * 10 + sum(1 for t in toks if t)
         if uniq <= 1 and width > 1:
@@ -43,6 +87,19 @@ def _pick_name_row(head, width):
     return best_i
 
 
+def read_metadata(head, k):
+    """Key = Value / Key: Value lines above the column row, kept for provenance."""
+    meta = {}
+    for row in head[:k]:
+        line = ", ".join(c for c in row if c).strip()
+        m = _PAIR.match(line)
+        if m and not _is_number(m.group("key")):
+            meta.setdefault(m.group("key").strip(), m.group("value"))
+        if len(meta) >= 40:
+            break
+    return meta
+
+
 def read_text(path):
     """Decode a data file, trying the encodings that actually occur in labs."""
     last = None
@@ -85,6 +142,10 @@ def sniff_delimiter(rows):
     width = max(len(g) for g in grid)
     if width >= 2:
         return None
+    if width == 1:
+        # "识别不出分隔符" would send the reader chasing encoding or comment
+        # prefixes; the real reason is that a single column cannot make a plot.
+        raise ValueError("文件只有一列数据：至少需要两列（x 和 y）才能解析成数据表")
     raise ValueError("识别不出分隔符，也没有多列空白对齐的数据")
 
 
@@ -122,16 +183,24 @@ def parse(path, comment=None):
         raise ValueError("找不到数值数据行：%s（只有表头？）" % path)
 
     k = _pick_name_row(head, width) if head else 0
-    longs = [head[k][c].strip() for c in range(width)] if head else [""] * width
+    raw_names = _row_cells(head[k], width) if head else [""] * width
     if k + 1 < len(head):
-        units = [head[k + 1][c].strip() for c in range(width)]
-    elif k > 0:
-        units = [head[k - 1][c].strip() for c in range(width)]
+        units = _row_cells(head[k + 1], width)
+    elif k > 0 and head[k - 1] and head[k - 1][0].strip().lower() in _UNIT_KEYWORDS:
+        units = _row_cells(head[k - 1], width)
     else:
         units = [""] * width
-    extra = len(head) - 2 if len(head) > 2 else max(0, len(head) - 1)
+    names = []
+    for c in range(width):
+        nm, un = split_name_units(raw_names[c])
+        names.append(nm)
+        if un and not units[c]:
+            units[c] = un
+    raw_names = names
+    metadata = read_metadata(head, k)
+    extra = k if metadata else max(0, len(head) - 2)
     if extra:
-        warnings.append("忽略了 %d 行额外表头（只取长名/单位两行）" % extra)
+        warnings.append("忽略了 %d 行文件说明（仪器参数已放进 metadata）" % extra)
 
     columns = []
     for c in range(width):
@@ -147,11 +216,11 @@ def parse(path, comment=None):
                 vals.append(None)
         live = sum(1 for v in vals if v is not None)
         if live == 0:
-            warnings.append("第 %d 列 %r 解析后全为空，已跳过" % (c + 1, longs[c] or "col%d" % (c + 1)))
+            warnings.append("第 %d 列 %r 解析后全为空，已跳过" % (c + 1, raw_names[c] or "col%d" % (c + 1)))
             continue
         columns.append({
             "index": c + 1,
-            "name": longs[c] or "col%d" % (c + 1),
+            "name": raw_names[c] or "col%d" % (c + 1),
             "units": units[c] or "",
             "values": vals,
             "numeric_rows": live,
@@ -165,5 +234,6 @@ def parse(path, comment=None):
         "rows": len(body),
         "header_rows": len(head),
         "columns": columns,
+        "metadata": metadata,
         "warnings": warnings,
     }

package/edge_checks.py

@@ -173,6 +173,16 @@ def main():
         print("SKIP  node launcher (node not on PATH)")
 
     bad = [n for n, ok in results if not ok]
+    # Same teardown as smoke_test: reclaim only the windowless instances this run
+    # started; an older one may belong to a live dsh session.
+    late = [r for r in S.originlab._process_rows()
+            if not r["window"] and str(r.get("started") or "") >= S.STARTED_AT]
+    for row in late:
+        S.originlab._close_pid(row["pid"])
+    left = [r for r in S.originlab._process_rows()
+            if not r["window"] and str(r.get("started") or "") >= S.STARTED_AT]
+    check("run leaves no Origin instance", not left, left)
+
     print("\n%d/%d passed" % (len(results) - len(bad), len(results)))
     if bad:
         print("FAILED:", ", ".join(bad))